Why Are My Cold Emails Going to Spam? (2026)

Cold email lands in spam for five reasons, and they are not equally important. In order of weight: the domain is too new or was never warmed, authentication does not actually pass, the list has enough bad addresses to push bounces above two percent, the volume jumped suddenly, or the complaint rate crossed 0.1 percent. Content is the fifth of five, which is why a well-written email from a cold domain still lands in spam.

This page is published by emailcampaign.ai, which sells cold email infrastructure. Every number here is one a provider publishes or one we send against, and the checks it recommends are free and need no account.

Work the causes in order, because the order is the answer

The mistake almost everyone makes is to start with the email. Rewriting the subject line is cheap and visible, so it is what people reach for first. It is also the variable a filter weighs least once your domain has a history, and it cannot move a domain that has no history at all.

Here is the order that actually resolves the problem, and roughly how long each takes to fix.

1. The domain is too new, or was never warmed

A filter decides where to put your message largely on what the sending mailbox has done before. A mailbox with no past has nothing for it to read, and a mailbox that sent four hundred messages on its first Tuesday has told it something worse than nothing.

Warm-up takes two to three weeks, from a handful of messages a day to the mailbox's working cap. During the ramp, warm-up traffic and campaign traffic share the same daily number, so a mailbox warming at eight a day has room for two campaign sends, not ten. That single confusion is the most common self-inflicted deliverability wound there is.

There is no honest shortcut. Pre-warmed mailboxes bought from a vendor carry someone else's sending history, usually on a domain you do not own and cannot take with you.

2. Authentication does not pass, even though the records exist

Most authentication problems are not missing records. They are records that exist, look right in the registrar dashboard, and fail anyway. Two causes account for most of them, and both produce a permanent error, which means the receiving system cannot evaluate SPF at all rather than evaluating it weakly.

The first is two SPF records on the same host. One arrived with the website host, another with the mail provider, and each is valid alone. Published together they are a permanent failure. The fix is a merge: one record, one v=spf1, every include inside it.

The second is an include chain past ten DNS lookups. Every include, a, mx, ptr, exists and redirect costs a lookup, includes nest, and the eleventh fails the record. A chain that worked with one provider breaks silently when you add a second.

There is a third that is specific to Microsoft. The DKIM CNAME target inside your tenant's onmicrosoft.com domain contains a per-tenant value, so a record copied from a guide or from another tenant resolves to nothing and DKIM quietly never signs. Read the exact target from the tenant you are configuring.

Check all of this against a public resolver rather than the registrar interface: the dashboard shows what you saved, and the provider sees what a resolver serves.

3. The list is pushing your bounce rate up

A hard bounce rate above about two percent reads to a provider as a bought or stale list, and it is the fastest way to lose a domain that has just finished three weeks of warm-up. Verify addresses before the first send rather than after the first bounce. Verification costs cents per address and buys back the only thing you cannot re-buy, which is the domain's history.

4. The volume pattern does not look human

Eight to ten cold emails per mailbox per day is the number that keeps a warmed mailbox healthy over months. Five to fifteen is the hard band. That is far below every provider's published ceiling, and deliberately so: Google allows two thousand messages a day and Exchange Online ten thousand recipients a day, but those numbers describe when the provider stops you, not when the provider starts distrusting you.

Spread sends across working hours rather than firing them in a block, and count follow-ups. The limit is total outbound per mailbox, so a three-step sequence reaches about a third as many new people as the raw send count suggests.

5. People are marking it as spam

Google publishes the threshold: keep the spam complaint rate below 0.1 percent and never let it reach 0.3 percent. At eight sends a day, one complaint takes a mailbox past 0.1 percent for that day, which is a useful way to feel how thin the margin is. Google Postmaster Tools shows the rate the filter actually sees for domains you send to Gmail, it is free, and it is the only place your reputation is described in Google's own words.

Then, and only then, the words

Content still decides three things. Older secure email gateways at large companies keep keyword weights and will quarantine a message carrying a pile of promotional vocabulary. Gmail's Promotions tab is content-driven, so an email that reads like an offer goes there even when it is authenticated and warm. And the reader, the only filter that replies, deletes on the first cliché.

That is worth five minutes at the end, not five hours at the start.

A diagnosis you can run in twenty minutes

Check the records against a public resolver and fix any permanent error. Read Google Postmaster Tools for the complaint rate and the domain reputation. Look at the shape of your refusals: a 5.1.1 means the address does not exist and says nothing about you, while a 5.7.x is a verdict about the sender. Count what each mailbox actually sends per day, warm-up included. Then run a placement test to seed mailboxes at Gmail and Outlook to see where mail is landing rather than guessing from open rates.